Table of contents
Introduction
We will explain the differences between the AI model conversion tools for ambaCV2X and ambaCV5X. The procedure for ambaCV2X is explained here. Please also check.
Differences in environment construction
Run CV Tool
Start the container.
$ sudo docker start [container name]
Log in to the container.
$ sudo docker exec -it -u cvtool [container name] /bin/bash
If you need a password, please use "cv".
If you want to run it for ambaCV5X, please do the following. It is not necessary when running for ambaCV2X.
Edit the /home/cvtool/setup_env.sh file to configure for ambaCV5X.
Edit the first line as shown below.(cv22 to cv52)
before
source /usr/local/amba-cv-tools-basic.2.5.2.0.1210.ubuntu-18.04/env/cv22.env
after
source /usr/local/amba-cv-tools-basic.2.5.2.0.1210.ubuntu-18.04/env/cv52.env
Set the environment (inside the container).
$ cd /home/cvtool $ source setup_env.sh
Differences between Caffe/Tensorflow/ONNX (PyTorch)
Setting.conf
CAVALRY_VER is "2.1.7" for ambaCV2X, but please set "2.2.8.2" for ambaCV5X.
(omission) #cavalry version #if not specified -> "" CAVALRY_VER="2.2.8.2" (omission)
This applies to all Caffe, Tensorflow, ONNX (PyTorch).